Personalizing web surfing with semantically enriched personal - PDF document

Personalizing web surfing with semantically enriched personal profiles Anupriya Ankolekar, Denny Vrandeˇ ci´ c Institute AIFB, University of Karlsruhe, Germany { anupriya,denny } @aifb.uni-karlsruhe.de Abstract. Personalization mechanisms on the web today are clumsy and obtrusive, because users need to log in to multiple websites and enter their personal information and preferences separately for each. In addition, the user profile is different for each website and cannot be combined with other information on the web. Using Semantic Web tech- nologies, in particular FOAF, we can identify the person browsing to the website. In this paper, we propose an extension of the HTTP GET method to include a new parameter that points to the URL of the user’s FOAF file. This simple, but powerful extension enables the web server to use information contained in the person’s FOAF file to personalize the web pages returned. We also present a proof-of-concept implementation by customizing our institute welcome page using a visitor’s FOAF file. 1 Introduction No one knows you are a dog on the Internet, or so the saying goes. While this may have its advantages, letting websites know who you are enables them to personalize your web surfing experience, serving you information customized for your needs and preferences. Knowing the interests, activities, acquaintances or accessibility problems of a user, the web site can adapt the web pages it serves and offer a personalized, familiar and welcoming experience for the user. The personalizations may include content personalization (show me the things I am interested in), or link personalization (show me only those links that I am likely to click on) and structure/navigation personalization (reorder the website as I am likely to understand and search through it). Several studies have explored the use of personalization and knowledge of people’s activities on a website to support web browsing and have found it to considerably enhance the browsing experience for users [15]. A key requirement for enabling such personalization is that the web server must be able to identify the person visiting the website and know her characteris- tics and preferences. Currently, the vast majority of sites attempt to personalize the user’s browsing experience by requiring the user to create an account on the website and login every time a personalized service is used. Amazon and Yahoo! , for example, follow this model. However, this is quite painful as the user needs to remember multiple logins and usually go through a log-in procedure at every site, which disturbs the smooth flow of web browsing. What we really want is a

seamless experience, where the person decides which profile she wants to adopt and then websites automatically recognize her and offer her customized content. Ideally, the user should not even be aware that she needs to identify herself to a website, but is naturally offered information relevant to her. What we need therefore is a user-side personal profile that describes a person and her interests, affiliations and acquaintances in an open standard. Semantic Web [2] technologies provide exactly this in the form of a FOAF (Friend of a Friend) [3] file. In this paper, we propose a personalization mechanism that allows a web server to identify a user’s FOAF file with every page request. The web server may then use the information in the FOAF file to personalize the page it returns, potentially retrieving additional information on the web in the process. This is where the advantages of using Semantic Web technologies become apparent. Since the FOAF file is in RDF [1], the information it contains can easily be combined with other RDF information on the web. In the following, we first discuss the concrete details of our approach (sec- tion 2) and then related work on personalization mechanisms (section 3). We then demonstrate the feasibility of our approach through a proof-of-concept imple- mentation in section 4, showing how the AIFB website personalizes its welcome page with information from a visitor’s FOAF file. Of course, this is still a fairly simple use of the information. More sophisticated usage of the data in the visi- tor’s FOAF file can enable a range of scenarios as discussed in section 5. Finally, we discuss some of the implications of our approach, in particular with respect to privacy in section 6, before summarizing the contributions of the paper in the concluding section 7. 2 Approach We rely on a user’s Friend Of A Friend (FOAF) [3] profile to personalize her browsing experience. The FOAF vocabulary is one of the most popular RDF vocabularies on the web and describes a person, in terms of several attributes of the person, such as homepages, affiliations, photographs and contact details, as well as specifying the acquaintances and friends of the person. Millions of FOAF files already exist, and are particularly popular within the blogging community. The FOAF vocabulary definitions are written as RDF [1] statements. This allows software to process the FOAF information and follow links to the FOAF files of friends and acquaintances to gather as much information as it needs. Using RDF as a description format has another advantage: we can easily combine FOAF statements with statements from other RDF vocabularies, such as RSS 1 for describing blog feeds, and the Web of Trust (WOT) Schema, to describe signatures on RDF documents, or geographical location vocabularies. Thus, a FOAF file can contain all kinds of information about the user and point to files with more specific information about the user. The web server can use as much of the information as it understands, ignoring the rest. So the system allows 1 http://en.wikipedia.org/wiki/RSS (protocol)

for a very graceful performance degradation, and still enables it to express any information that could be useful for personalization. To give web servers access to a user’s FOAF profile, we extend HTTP’s GET method to include a new parameter that points to the URL of the user’s FOAF file. When the web server receives a GET request, it examines its parameters in order to determine which information to send back. At the most obvious level, this includes the web page to return. However, there is also an element of browser-personalization involved. For example, depending on which browser is requesting the page, Internet Explorer or a Mozilla browser, the web server can present different versions of the web page best suited to the browser. By extending the GET parameter to point to a user’s FOAF file, the web server can fetch information about the user and customize the returned page for the user. This is a fairly simple idea, but it is attractive for several reasons: 1. By relying on the HTTP protocol, we are using the lowest common denomi- nator when it comes to web access, so this method can be used with minimal modifications by most websites. 2. This eliminates the need for the user to login to a site explicitly, with all its associated problems of remembering the login/password details, the time and effort required for the login procedure as well as the break in the smooth navigation through the web. 3. Personal information is within the control of the user rather than multiple websites, so the user can decide what information to expose. Unlike previ- ous user passport methods [12], the FOAF files are simple to understand, straightforward to create, and based on an open standard format. 4. The FOAF file is passed with every HTTP GET call, enabling the user to change profiles within a given session. 3 Related work There are several mechanisms used currently for achieving personalization on the web [9]. These include server-side accounts , which require the user to create an account on the website and log in to it when making use of personalized services, cookies , used for storing identification and user preferences on the user’s machine, and identity profiles , such as Microsoft Passport [12], AOL Screen Name 2 or OASIS Open Identity , which provide a single sign-on for multiple services. Server-side accounts are provided by most major websites and portals such as Amazon and Google , and cookies are used (often indiscriminately) by an even larger number of websites. However, there are distinct disadvantages to all of these approaches. To begin with, none of these mechanisms give fine-grained control to the user of the information he or she is presenting to a website. Server- side accounts usually require a standard list of information regardless of whether and how it is used subsequently. Furthermore, there is no single point of control for the user. With server-side accounts, user information is scattered all over 2 https://my.screenname.aol.com/